AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

Hello from an organiser of the Loebner Prize 2016
 
 
  [ # 16 ]

Well observed Denis…


I received four typos in mine:

“whats yours” on question #1,
“whats up”, on question #2,
“whats my name” on question #3,
“cimb” on question #7


I also have to add that the typos are not linked to the transcript but to the question received by the bot itself (all same answers successfully checked with a copy of my program). So there was a mistake here.

That’s unfortunate of course. In conversation I expect any program to have to “beat it”, and my program did not beat that usual typo, so there is a mistake of mine.

But in any case, I would hate to be a frustrated guy on that one… but I believe on ranking questions we should receive the same sentences and I think all would agree.

I am sure this was unintentional and I believe this may be because the input on the raspberry or bluetooth keyboard got mixed up.

I believe entrants who received typos should ask for a short rerun of the questions which had typos. New post coming with the differences I had.

 

 

 
  [ # 17 ]

Some clarifications now about the differences I have:

”...Andrew, whats yours” -> memory got totally mixed up even though my bot replied cautiously.
“whats up” -> got unanswered.
“whats my name” -> got unanswered.
“cimb” -> was just not understood and mixed the answer.

Memory got totally mixed up from question #1, saving identity as “Andrew whats” or something like it and a few questions might be broken even if re-asked.

However Andrew, for a rerun you might still ask the bot:

“what’s my name”
(and see answer from #1, but maybe broken)

“the cat tried to climb in the box but got stuck because it was too big. What was too big?”
(you’ll get: “The climb”, which actually is funnier, after all the cat may be inside the box when climbing and stay stuck inside the box? wink well ... )

There does not seem to be any difference with “whats up”.

So there shall be only a few points taken.
All entrants shall still feel safe, because I don’t think the points I’ll get won’t make me a finalist.
Isabelle got an average to low performance this year.

In am still very interested in knowing whether my position would be better!

 

 

 
  [ # 18 ]

@ Andrew
Sent you an email with instructions about two of the questions that could be safely typed again with no typo. Others cannot now, that would break contextual memory.

@ All
Don’t worry guys, I am sure all finals will stay final wink

 

 
  [ # 19 ]

For what it’s worth: When I ran contest, I had great difficulty typing in screening questions without making typos (actually it was impossible), so I wrote a small (~ 25 lines) Perl program that read a file with the questions one question per line and output each line (ie question by question) in lpp format.  My testing program also wrote the questions and the responses to a text file so the programs could be judged.

The first time I used the program it didn’t issue a line feed ( carriage return?) after each question, although each question did end with an interrogative “?”.  Some of the entries couldn’t deal with it and didn’t respond.

The entries that didn’t respond were immediately eliminated, which made the screening process sooo much easier.  I didn’t have to think.

 

 
  [ # 20 ]

Thanks Hugh!
I also think we are not in anything like an exceptional situation here.

Typo itself is minor, common speech (slightly less for the “Mt” that hit Johnny), and something a good program should really have managed (meaning any program should survive a “conversation” which is much worse regarding typos!).

I believe we are more in a very common minor mishap situation in that all programs did not receive the same treatment for selection. In any case, AISB will rule it the way it wants to, and we’ll at best end up with scores fixed slightly and no change on the big picture.

I am still very proud of what the AISB did here and don’t want to make it sound bad in any way.
Still makes me want to live in England smile

 

 
  [ # 21 ]

I agree. If there are going to be typos, everyone should get the same or none. There have been typos in the qualifying round every year that I’ve entered and it’s about time it gets automated.

On a different subject: How much idle time is there usually between rounds?
I ask because I’ll need some way to distinguish “Hello?” the greeting, from “Hello?” the reminder, since the programs aren’t physically around to hear new rounds announced, and judges don’t always say goodbye at the end of one.

 

 
  [ # 22 ]

There is officially 5 minutes between rounds. However, the bots are usually restarted after each round so they can tell when a new judge is talking to them or keep a count of which round it is.

Will you be attending? It would be good to meet up in real life.

 

 
  [ # 23 ]

Restarting would solve that issue, as long as they also remember to reset the LPP.
I haven’t decided whether to attend yet. I can get to London by train and I can walk, but subway travel and hotel arrangements may be a little beyond my managing skills, if customs even still allow Europeans into Brexit tongue rolleye. Would be nice to meet up but I don’t expect to look into it until the last week. Know any nearby sleeping accommodations?

 

 
  [ # 24 ]

Not sure what you mean by restart the LPP but the organisers restart the judge program and the bots between rounds. Otherwise, fragments of the previous rounds may appear for the judges and your bot may also think it is talkin gto the previoss judge.

As for travel, I stayed at this one when I went in 2014:
https://www.travelodge.co.uk/hotels/468/Milton-Keynes-Shenley-Church-End-hotel

It was ok for a night but not quite the Ritz smile

 

 
  [ # 25 ]

Thanks. Yes, I mean the judge program. Thing is, if you restart my program it’ll start outputting numbered characters from #1 again, and if the judge program left off at character #99901 and is not reset, the judge program doesn’t pick up on any output sent. I believe that’s what happened to Bruce one particular round in the past.

 

 
  [ # 26 ]

Let me point out that in fact rose was NOT restarted between the 3rd and 4th round in the year before. Leading to confusion because she was continuing an old conversation with a new judge

 

 
  [ # 27 ]

Good to know. That’s more in line with my expectations.

 

 
  [ # 28 ]

It was the same year that one of the “judges” merely walks away and doesnt want to talk to chatbots, so 2 of Rose’s 4 rounds were “contaminated”

 

 
  [ # 29 ]

A shame that happened. The entire contest should have been declared VOID at that point and either rescheduled or restarted, to my twisted way of thinking.

Seems that the pressure is not on the judges so there’s no consequence for walking away or not playing fairly. Pity that.

 

 
  [ # 30 ]

I have an entire paper on “winning the loebners” which recounts my experiences…
https://sourceforge.net/projects/chatscript/files/?source=navbar

paper: Winning the loebners.

 

 < 1 2 3 > 
2 of 3
 
  login or register to react